Apply for a GPU community grant: Personal project

#1
by hiro1997 - opened

Subject : Free GPU Upgrade Request for Open-Source Basketball Highlight AI

Description :

Hi Hugging Face team,

I'm building an open-source basketball highlight detection tool that automatically identifies exciting moments in street basketball videos and generates cinematic highlight reels. The project uses YOLOv8-Pose for player detection, X-CLIP for semantic understanding, and custom rendering for visual effects.

The project is fully open-source: https://github.com/hirochen1997/basketball_highlight

Currently on CPU Basic, a 10-minute video takes 30+ minutes to analyze, making the demo practically unusable for visitors. With a T4 GPU, the same analysis would complete in ~7 minutes, providing a much better user experience.

GPU acceleration is essential because:

  • YOLOv8-Pose inference is ~10x faster on GPU
  • X-CLIP video encoding requires significant GPU memory and compute
  • Optical flow and scene detection benefit greatly from CUDA acceleration
    This is a community project for amateur basketball players to create shareable highlights from their pickup game videos. A free GPU upgrade would make this demo accessible and usable for the community.

Thank you for considering this request!

Sign up or log in to comment